October 33, 207
Serge Lang lecture This year s Serge Lang Undergraduate Lecture will be given by Keith Devlin of Stanford University. The title is When the precision of mathematics meets the messiness of the world of people. The lecture will be given at 4:0PM today in 60 Evans. Please visit the Facebook event page for the lecture as well.
Dinners Foothill DC dinner, Friday (Nov. 3) at 6:30PM. Clark Kerr DC, Sunday (Nov. 5) at 6PM. Yuge crowds at both dinners, OK? We want to have the biggest crowds ever. So come.
Recall the fundamental concepts Probability space Random variable Probability density function (PDF) Cumulative distribution function (CDF)
Computing averages Take a (biased) coin that comes up heads () 3/4 of the time and tails (0) /4 of the time. Flip the coin three times: Ω = { 000, 00, 00, 0, 00, 0, 0, }, P(000) = 4 4 4, P(00) = 4 4 3 4, etc. We do not have a uniform probability space!
Random variable As on Tuesday, let X : Ω { 0,, 2, 3 } be the random variable that takes each string in Ω to the number of s in the string: X(000) = 0, X(00) =, X() = 3, etc.
Mean What is the average value of X? The average value is called the mean BTW. In order to compute the mean, we have to say what we mean by the mean.
Mean What is the average value of X? The average value is called the mean BTW. In order to compute the mean, we have to say what we mean by the mean.
Mean What is the average value of X? The average value is called the mean BTW. In order to compute the mean, we have to say what we mean by the mean.
Mean s meaning, first try We sum over all strings and weight each string by its probability. For each string we compute X(the string). We add the result: µ = X(000)P(000) + X(00)P(00) + = 0 4 3 + 3 4 3 +. There are eight terms in the sum.
µ = X(000)P(000) + X(00)P(00) + = 0 4 3 + 3 4 3 +. There are eight terms in the sum, one where X = 0, three where X =, three where X = 2, one where X = 3. The strings for which X = (for example) are: 00, 00 and 00. Their contribution to µ is P(00) + P(00) + P(00), which we rewrite as P(X = ).
Mean s meaning, second try µ = 0 P(X = 0) + P(X = ) + 2 P(X = 2) + 3 P(X = 3) = i i P(X = i) = i i f (i). In the last two equations, i ranges over the possible values of X, namely 0,, 2, 3 and 4. Also, f has the same meaning as in yesterday s discussion of coin tosses: it s the Math 0B analogue of the probability density function. Namely, f (i) = P(X = i).
Mean s meaning In the Math 0B world, if a random variable is distributed according to f, then mean of X = x x f (x), where x runs over all possible values of X. We could list the possible values of X: x, x 2,..., x k and let Then p i = f (x i ) = P(X = x i ). µ = k p i x i. i= This is how Schreiber writes the mean.
Another example What is the mean height of students in this class? Say Ω = the set of students in this class and think of Ω as a uniform probability space. Let X : Ω { 20, 2, 22,..., 200 } be the height function, where heights are measured in inches and rounded up or down to the nearest inch. The first approach to computing the mean height is to sum up everyone s height and to divide by the number of students in the class 235, let s say.
The approach taken by the second formula (two slides ago) is to count the number of students with a given X-value, say X = 65. Imagine that 2 students have height 65; then P(X = 65) = 2. Once we have tabulated the number of 235 students with each possible height, we can compute the mean 2 very easily by adding up numbers like 65 235.
Expected value According to Wikipedia: In probability theory, the expected value of a random variable, intuitively, is the long-run average value of repetitions of the experiment it represents.... The expected value is also known as the expectation, mathematical expectation, EV, average, mean value, mean, or first moment. In other words: They are just synonyms. Expected value = mean.
Expected value Consider the probability space consisting of the two possible outcomes of the flip of a fair coin: Ω = { 0 }, each outcome occurring with probability 2. Let X(0) = 0, X() =. What is the expected value of X? Answer: µ = 0 P(X = 0) + P(X = ) = 0 2 + 2 = 2.
Expected value Consider the probability space consisting of the two possible outcomes of the flip of a fair coin: Ω = { 0 }, each outcome occurring with probability 2. Let X(0) = 0, X() =. What is the expected value of X? Answer: µ = 0 P(X = 0) + P(X = ) = 0 2 + 2 = 2.
Expected value The expected value 2 of X is not a value of X. We do not expect the expected value of X because it is not a value of X. The terminology expected value is misleading and therefore bad.
Back to Regrades have been enabled for MT#2. Please do not request a regrade on a problem unless you have already discussed your solution with your GSI. Your request should contain a statement along the lines of My GSI Ken Ribet said this was okay.
Back to If X has PDF equal to f, then mean of X = x f (x) dx.
Pareto Fix p > and note that The function is then a PDF. x p dx = p What is the mean of this PDF? ] x p = p. 0 if x <, f (x) = p x p if x
Pareto The mean in this case is The integral x p x p dx = (p ) x q dx converges only for q >, when its value is x p dx. q. Hence a Pareto-distributed X has finite mean only for p > 2. The mean in this case is p p 2.
Comparisons When we work with PDFs and CDFs, and when we compute means, we deal with improper integrals like dx. This x p integral has the form where g(x) is positive. a g(x) dx, An integral like this is either finite (convergent) or infinite (divergent).
Consequently, if h(x) g(x) and both h and g are positive, we have a g(x) dx < = a h(x) dx <. If the integral of a big function g is convergent, so is the integral of a smaller function. That s comparison for you. Logically, this means that if the integral of the smaller function is infinite, then so is the integral of the larger function: a h(x) dx = = a g(x) dx =.
HW Example The integral Consequently, + e x > e x, so g(x) = e x.... 0 e x dx is easily evaluated: it s, so it s finite. dx is convergent: indeed, 0 + ex + e x < e x = e x. We take h(x) = + e x, This is 7.2 #, but I changed the lower limit of integration from to 0.
HW Example The integral Consequently, + e x > e x, so g(x) = e x.... 0 e x dx is easily evaluated: it s, so it s finite. dx is convergent: indeed, 0 + ex + e x < e x = e x. We take h(x) = + e x, This is 7.2 #, but I changed the lower limit of integration from to 0.
Want to know more? An antiderivative of + e x turns out to be x ln( + ex ); this will be easy to check on the doc camera. Consequently, 0 + e x dx = (x ln( + ex )) ] = lim x ((x ln( + e x )) + ln(2). The limit is 0 because e the limit = by l Hôpital s rule. Hence the value of the integral is ln 2 0.693. It is validating that 0.693 is less than. 0
A good limit problem Find the limit as x of the difference x ln( + e x ). This is a quintessential example.
The integral test While discussing comparisons, we should highlight this fact: Take a decreasing function f (x) on [, ) with the property that lim f (x) = 0. Typical examples: f (x) =, f (x) = x x x 2 ; more generally, f (x) = with p > 0. Then x p f () + f (2) + f (3) + < f (x) dx <.
Example For p > 0, the series n= In particular, the harmonic series diverges; we saw that before. converges if and only if p >. np + 2 + 3 + 4 +
Why is there a comparison between series and integrals? This is based on the diagrams that we drew for left- and right-endpoint approximations to integrals. I ll redraw a few diagrams on the document camera.
Come to dinner Dinner tonight in the math department after Keith Devlin s talk. Dinner on Friday at Foothill DC (6:30PM). Dinner on Sunday at CKC DC (6PM). Let s do this, Bears!